Research map / Graph learning / GNN architectures
Towards Deepening Graph Neural Networks: A GNTK-based Optimization Perspective
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use Critical DropEdge instead of the strongest listed baseline | deep GNNs · Cora, 32-layer | test accuracy (%): 82.1 \pm 0.5 -> 82.6 \pm 0.9 | Helped |
| Use Critical DropEdge instead of the strongest listed baseline | deep GNNs · Citeseer, 32-layer | test accuracy (%): 70.2 \pm 0.8 -> 70.7 \pm 0.9 | Helped |
| Use Critical DropEdge instead of the strongest listed baseline | deep GNNs · Pubmed, 32-layer | test accuracy (%): 77.2 \pm 1.3 -> 77.8 \pm 1.0 | Helped |
| Use Critical DropEdge instead of the strongest listed baseline | deep GNNs · Physics, 32-layer | test accuracy (%): 91.3 \pm 0.8 -> 91.6 \pm 0.6 | Helped |
| Add residual connection to aggregation (ablation) | wide GCNs · large-depth GNTK convergence | convergence rate | Helped |
| Add residual connection to MLP transformations (ablation) | wide GCNs · large-depth GNTK convergence | convergence rate | No clear effect |
| Add residual connections to both aggregation and transformations (ablation) | wide GCNs · large-depth GNTK convergence | convergence rate | Helped |
| Vary the edge preserving percentage instead of fixing it at the critical rate (ablation) | C-DropEdge · Cora, Citeseer, and Pubmed | test accuracy (%) | Mixed |
Closest papers by meaning
- Old can be Gold: Better Gradient Flow can Make Vanilla-GCNs Great Again (NeurIPS 2022 · similarity 0.93)
- On Provable Benefits of Depth in Training Graph Convolutional Networks (NeurIPS 2021 · similarity 0.92)
- Generalization Guarantee of Training Graph Convolutional Networks with Graph Topology Sampling (ICML 2022 · similarity 0.92)
- How Graph Neural Networks Learn: Lessons from Training Dynamics (ICML 2024 · similarity 0.90)
- On Vanishing Gradients, Over-Smoothing, and Over-Squashing in GNNs: Bridging Recurrent and Graph Learning (NeurIPS 2025 · similarity 0.89)
- A Manifold Perspective on the Statistical Generalization of Graph Neural Networks (ICML 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.